Abstract
Purpose :
Since 2016, the integration of artificial intelligence and deep learning in ophthalmology has accelerated, leading to the formation of many startups. This review assesses the development of AI technologies in ophthalmology, tracing their journey from their initial technology advancements to the formation of startups or partnerships with existing companies, and then subsequently, their attainment of regulatory approvals, and the range of eye diseases addressed by these innovations.
Methods :
A systematic review was conducted to identify AI technologies in ophthalmology from 2016 onwards. We examined academic papers discussing AI ophthalmology technologies. Recognizing that not all relevant information would be found in academic publications, the search was extended to FDA's databases using specific terms such as 'PIB' and 'NFJ, and EUDAMED, to identify AI technologies and companies pertinent to ophthalmology. The eye diseases targeted by these technologies were also identified.
Identified the AI technologies that led to the establishment of startups or collaborations with existing companies by cross-referencing their technological origins with records of company formations and partnership announcements. To validate their regulatory approvals, the respective databases were consulted (FDA, CE, TGA, HSA, etc.) and Google searches for unlisted startups.
Results :
Since 2016, a total of 47 AI algorithms in the field of ophthalmology have been developed. Of these, 81% (n=38) have transitioned into startups or acquired by companies, with 68% (n=26) successfully receiving regulatory approval. This development has been steadily increasing, particularly with a notable spike in 2022, likely influenced by the 2021 approval of CPT code 92229. 58% of these startups received CE IIa approval, with only 24% securing FDA approval, suggesting a preference for the European market.
In terms of disease focus, a significant majority of the AI algorithms (94%) were developed for diabetic retinopathy, and 38% targeted age-related macular degeneration. Additionally, 51% were designed to address multiple ophthalmic diseases.
Conclusions :
The rapid growth of AI-driven ophthalmology, particularly in diabetic retinopathy and multi-disease applications, reflects a transformative shift in eye care. AI technologies becoming startups and obtaining regulatory approvals suggests promising significant advancements in eye care practices and research.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.